Find the unique polynomial of degree 2 or less which fits the following data:
x:0,1,3f(x):1,3,55
Also obtain the bound on the truncation error.
Technique
Use Newton’s divided-difference interpolation on the three unequally spaced points to get the unique polynomial of degree ≤2. The truncation error uses the standard remainder f(x)−P2(x)=3!f′′′(ξ)ω(x) with ω(x)=(x−x0)(x−x1)(x−x2).
Solution
Data points: (x0,f0)=(0,1), (x1,f1)=(1,3), (x2,f2)=(3,55).
Newton’s divided differences
First order:
f[x0,x1]=1−03−1=2,f[x1,x2]=3−155−3=252=26.
Second order:
f[x0,x1,x2]=3−026−2=324=8.
For interpolation at three nodes, the error at a point x is
E(x)=f(x)−P2(x)=3!f′′′(ξ)ω(x),ω(x)=x(x−1)(x−3),
for some ξ in the smallest interval containing {0,1,3,x}. Hence
∣E(x)∣≤6M3∣ω(x)∣,M3=max[0,3]∣f′′′(ξ)∣.
To bound ∣ω(x)∣ on [0,3]: ω′(x)=3x2−8x+3=0 gives x=68±64−36=34±7, i.e. x≈0.4514,2.2153. Evaluating:
∣ω(0.4514)∣≈0.6311,∣ω(2.2153)∣≈2.1126,
and ω=0 at the nodes. So
max[0,3]∣ω(x)∣=∣ω(2.2153)∣≈2.1126(exactly at x=34+7).
Therefore the uniform bound on the truncation error over [0,3] is
∣f(x)−P2(x)∣≤6M3[0,3]max∣ω(x)∣≈6M3(2.1126)≈0.3521M3,M3=[0,3]max∣f′′′∣.
(If the underlying f is itself a polynomial of degree ≤2, then f′′′≡0 and the interpolation is exact.)
Answer
P2(x)=8x2−6x+1. Truncation error E(x)=6f′′′(ξ)x(x−1)(x−3), bounded on [0,3] by 6M3max∣ω∣≈0.3521M3, the maximum of ∣ω∣ occurring at x=34+7.
We've mapped all 13 years of this exam. Get new solutions, tools, and guides as we release them — free.